Entropic equation of state and scaling functions near the critical point in scale-free networks

Entropic equation of state and scaling functions near the critical point in scale-free

networks

C. von Ferber, 1 - 2 ^ R- Folk, 3 'Q Yu. Holovatch, 4 ' 3 '! R. Kenna, 1 '! and V. Palchykov 4 -0

'Applied Mathematics Research Centre, Coventry University, Coventry CV1 5FB, UK
2 Physikalisches Institut Universitat Freiburg, D-79104 Freiburg, Germany
3 Institut fur Theoretische Physik, Johannes Kepler Universitat Linz, A-4040, Lmz, Austria
^Institute for Condensed Matter Physics, National Academy of Sciences of Ukraine, UA-79011 Lviv, Ukraine

(Dated: January 19, 2011)

We analyze the entropic equation of state for a many-particle interacting system in a scale-free
network. The analysis is performed in terms of scaling functions which are of fundamental interest in
the theory of critical phenomena and have previously been theoretically and experimentally explored
in the context of various magnetic, fluid, and superconducting systems in two and three dimensions.
Here, we obtain general scaling functions for the entropy, the constant-field heat capacity, and the
isothermal magnetocaloric coefficient near the critical point in scale-free networks, where the node-
degree distribution exponent A appears to be a global variable and plays a crucial role, similar to
the dimensionality d for systems on lattices. This extends the principle of universality to systems on
scale-free networks and allows quantification of the impact of fluctuations in the network structure
on critical behavior.

PACS numbers: 64.60.aq, 64.60.F-, 75.10.-b

I. INTRODUCTION

Phase transitions and critical behavior in complex net-
works currently attract much attention [l[ because of
their unusual features and broad array of applications,
ranging from socio- [H to nanophysics Q. It is by now
well established that the critical behavior of a many-
particle interacting system located on the nodes of a gen-
eral network may crucially differ from that of a system
located on the sites of a d-dimensional regular lattice. Of
particular interest are the so-called scale-free networks,
for which the probability to find a node of degree k (i.e.
with k nearest neighbors) vanishes for large k as a power
law

P(k)

(1)

The questions we address in this paper concern two
fundamental principles of critical phenomena: universal-
ity and scaling Q. Both of these questions have to be
reconsidered when a system resides on a network. Usu-
ally, the universality of critical phenomena is understood
as stating that the thermodynamic properties near the
critical point T c are governed by a small number of global
features, such as dimensionality, symmetry, and the type
of interaction. In turn, the scaling hypothesis states that
the singular part of a thermodynamic potential near T c
has the form of a generalized homogeneous function. To

* C . vonFcrbcr@coventry. ac. uk

1 rcinhard.folk@jku.at

1 hol@icm p. lviv. ua|

- R.Kcnna@covcntry.ac.uk

^ palchykov@ic mp.lviv.uaf

be specific, for the Hclmholtz potential of a magnetic
system the latter can be written as Q

F(T,M)=T 2 -*f ± (M/Tp),

(2)

where M is the magnetization, r = |T — T c \/T c , a, f3 are
the universal exponents, and the sign ± corresponds to
T > T c or T < T c respectively. The essence of relation ([2])
is that the two- variable function F{t,M), when appro-
priately rescaled, is expressed in terms of a single scaling
variable leading to the scaling function f±{x). The ex-
pression ([2]) gives an example of the scaling function for
the Helmholtz potential F(t,M). Together with other
scaling functions - for the equation of state and for ther-
modynamic functions - this appears to give a suitable and
comprehensive description of critical phenomena @, H|.
These scaling functions are also universal in the sense
explained above.

For systems on scale- free networks the principle of uni-
versality is extended: there, the node-degree distribution
exponent A in Eq. (JT|) appears to be a global variable and
plays a crucial role, similar to the dimension d for systems
on lattices (see e.g. [9141^). All systems that belong to a
given universality class are governed by the same values
of critical exponents and critical amplitude ratios and
share the same universal form of scaling functions. Re-
cently, the scaling function formalism has been applied to
describe the critical behavior of magnetic systems with
the structure of a scale- free network 14|. There, scaling

functions for the magnetic equation of state and isother-
mal susceptibility were derived. In this paper we are
interested in the entropic form of the equation of state.
In particular this opens the possibility to derive scaling
functions for the heat capacity. These are of wide and
fundamental interest in the theory of critical phenom-
ena and have been the subject of thorough theoretical
and experimental studies for various magnetic, fluid, and

2

superconducting systems jl6H23|. A particular point of
interest concerns the existence or non-existence of the
moments of the distribution (fT]) on the universal behav-
ior.

The set-up of the the paper is as follows: in section
HI! we define the notations and derive expressions for
the entropic equation of state, heat capacity and mag-
netocaloric coefficient scaling functions for systems on
scale-free networks. These expressions are further an-
alyzed and compared with corresponding functions for
bulk systems in section UIT1 The paper concludes with a
summary and an outlook in section IIVI

II. FREE ENERGY AND SCALING
FUNCTIONS

The critical behavior of a many-particle system inter-
acting on a scale-free network crucially depends on the
node degree distribution P(k) via the decay exponent A
in Eq.([T]) [lj. In particular, for an infinite network the
value of A determines the order of the first diverging mo-
ment, this order being the lowest integer j > X—l. This
is reflected by the phase transition scenario. For low val-
ues of A < 3 the system remains ordered for any finite
temperature, whereas for A > 3 a finite temperature,
order-disorder phase transition occurs. Moreover, critical
exponents that govern a second-order phase transition in
a scale-free network attain their usual mean-field values
for high A > 5 and demonstrate non-trivial A-dependence
in the region 3 < A < 5. Logarithmic corrections to scal-
ing laws appear at A = 5: this resembles phenomena that
occur at marginal space or order parameter dimensions
in bulk systems [24[ .

A starting point for our analysis will be the expression
for the free energy of a system with scalar order param-
eter on a scale-free network. To be specific, from now on
we will consider ferromagnetic ordering and the sponta-
neous magnetization M as the order parameter with the
conjugate magnetic field H. In this case the correspond-
ing microscopic degrees of freedom are the Ising spins.
However, generalization to models of more complicated
symmetry is straightforward [HI, [l3|]. Due to the fact
that the type of networks under discussion are assumed
to have a local tree-like structure, the mean-field approx-
imation is asymptotically exact in the sense that thermal
fluctuations can be neglected. This leads to a form of the
free energy also found by other techniques. The lowest
order contributions to the singular part of the Helmholtz
free energy in the vicinity of T c are 0-[H|

F{M,T) = ^(T-T C )M 2 + ^M 4 , A > 5, (3)

F(M, T) = —(T— T C )M 2 + j-M" A_1 , 3 < A < 5. (4)

The parameters a, b > and the critical temperature
T c are A-dependcnt. This dependence can be made ex-
plicit using microscopic approaches 0, [1(1 [13 or may be

a

P

7 S

LJ

Ctc 7c

A > 5

1/2

1 3

1/2

2/3

1/3

3<A<5 £=§

i

A-3

1 A-2

A-4
A-3

A-5 A-3
A-2 A-2

A-4
A-2

TABLE I: Critical exponents governing temperature and field
dependencies of the thermodynamic quantities for different
values of A.

postulated in a Landau-like approach [TJ U|. For the
subsequent analysis we will absorb the parameters into
the dimensions of the corresponding obscrvablcs, passing
to dimensionlcss quantities,

r 1

/(m,r) = ±-m 2 + -m 4 , A > 5, (5)

/(m,r) = iV + ^m^ 1 , 3 < A < 5, (6)

with obvious relations between dimension-dependent and
dimensionless variables,

m = M/M , r = \T-T c \/T c , f = F/F , (7)

where M A ~ 3 = aT c /b and F = aT c M$. Since r mea-
sures the absolute distance to the critical point, the free
energy has two branches, corresponding to signs '+' and
'-' in Eqs. © and © for T > T c and T < T c , respec-
tively. It is easy to verify, that a system with the free
energy (J5]) , (J6]) possesses a second-order phase transition
at r = 0. Here, we employ the standard notation for
critical exponents governing the temperature- and field-
dependencies of the thermodynamic functions. For h =
and T — > , these are

Cfc-^r"", xt^^t-^ m T ^B±T- u (8)

while for T — > T~ one also has

m ~ Bt^ . (9)

On the other hand, for r = 0, the standard definitions
are

c h ~ A c h~ a % h~D c m\m\ s -\ (10)
XT ^ T c h^, m T ~B c T h-^. (11)

(See section III CI for the definition of the magnetocaloric
coefficient tut-) The values of these critical exponents
arc summarized in Table U [^-[H, (3 ■ It is worth noting
here that one way to derive the listed exponents is to
consider the naive dimensions of different terms in the
Landau free energy, similar to the standard field theoret-
ical procedure (see e.g. [HI)- With the values of critical
exponents to hand, one can rewrite the singular part of
the Helmholtz potential in the form of a generalized ho-
mogeneous function © 0:

/(m,r) = T 2 f±(x), x^m/r^, A > 5, (12)
/(m,r) = r^=3/±(x), x = m/r^, 3 < A < 5, (13)

3

where the free energy scaling functions are given by \\A

with function Ch given by

/±(s)

1 9

±-x
1

±-x 2 + -x x - 1
4

A > 5, (14)
3<A<5. (15)

Assuming that the Helmholtz potential is a complete dif-
ferential

dF

-SdT + HdM

(16)

one can further proceed with an analysis based on the
magnetic form of the equation of state,

H

dF
dM

or entropic form of the equation of state [25|

b - or

M

(17)

(18)

As we have noted in the introduction, the scaling func-
tions for the magnetic equation of state (both in the
Widom-Grifhth pj and Stanley-Hankey [f| forms) and
isothermal susceptibility have recently been reported
elsewhere (l4| . Here, we will proceed by analyzing the
entropic equation of state (|18p and heat capacity scaling
functions.

In terms of dimcnsionless variables Eqs. (fl~7|) . (fT8f take
on the form

h(m, t)

df{m,T)

dm

s(to,t) = =F

df(m,r)

(h

(19)

with field h and entropy s measured in units of Fq/Mq
and Fq/T c correspondingly. As before, and throughout,
the index ± refers to temperatures above and below the
critical point T c .

Since the free energy ([3]), (0| is explicitly a linear func-
tion of t, one obtains the usual mean field result for the
heat capacity at constant magnetization:

r -r ds

= 0.

M

(20)

To find the dimensionless constant-magnetic-field heat
capacity (25j .

Ch = ±

T 9s(r,m)

T c dr

(r±l)

ds (r, to)
dr

(21)

one can consider the entropy as a function of magnetic
field and temperature s(r, m(r, h)) which leads to

(r±l)

" ds

ds

9to(t, h)

.dr

m

dm

t dr

h-

that

ds/dm\ T

(22)

ds/dr\ m = one finally arrives at the expression for
the heat capacity,

Ch(T,m) = =pm

dm(r, h)

dr

(24)

Let us now consider separately the cases of fast (A >
5) and slower (3 < A < 5) decay of the node degree
distribution (TTJ).

A. A > 5

The free energy (0J leads to the expression for the en-
tropy,

TO

~2"

(25)

s(t, to)

which can be easily recast in a scaling form

s(T,m) = rS(x), (26)

where the scaling variable x = to/t' 3 = to/t 1 / 2 and the
entropy scaling function S(x) is

(27)

To obtain the heat capacity (|22|) we first write the mag-
netic equation of state ([TTJt

h = ±tto + m 3 (28)

and differentiate it with respect to r to obtain:

dm

±r + 3to 2

(29)

Substituting this into (|24[) leads to the representation of
Ch in the form of a generalized homogeneous function,

Ch(r, m) = C±(x),

(30)

with the scaling variable x defined above and the heat
capacity scaling function

C±(x)

2>x 2 ± 1 '

(31)

Note, that in (|3H|) the heat capacity exponent vanishes,
a = 0.

B. 3 < A < 5

A particular feature of the entropy of a system on a
scale-free network is that its dependence on magnetiza-
tion both for 3 < A < 5 and for A > 5 is given by Eq.
(|25| . In terms of the scaling function for 3 < A < 5 it
reads

c h = (1 ± T)C h {r,m)

(23)

,{t,x)=t 2 /^S{x),

(32)

4

where the entropy scaling function does not change and
is given by Eq. ([27)1 . The power of r is equal to 1 — a
and the scaling variable is now

m/T^m/r 1 /^.

(33)

However, the magnetic equation of state (fT!?]) for the
Helmholtz function ((6]) becomes A-dcpcndcnt:

irffl

A- 1

A-2

(34)

As in the previous subsection we obtain from this the
derivative 9m/9r|/j, and by substitution into Eq. (|24[) we
arrive at the representation of Cu in the form of the gen-
eralized homogeneous function

C h (T,m)

'C±(a;)

(35)

where the scaling variable x is given by (|33l) and the
heat capacity scaling function attains a non-trivial A-
dependence

C±(x)

x~

(A-l)(A-2) A

A ^

± 1

(36)

Note that on the basis of the scaling functions S(x)
in Eq. ([2Tjl and C±(x) in Eqs.([3i"]) and ([36]), one easily
obtains the corresponding scaling functions with respect
to the rescaled magnetic field

y = h/r^ 5 .

The connection between the variables x
from the magnetic equations of state
is given by

(37)

and y results
and ([Ml), and

y = ±x + x ,

A-l

V = ^ x H — :

.A-2

A > 5,
3 < A < 5.

(38)
(39)

Solving the above equations with respect to x and sub-
stituting the result x(y) into the functions S(x) and
C±(x) leads to the scaling functions S(y) and C±(y).
The behavior of the above scaling functions will be an-
alyzed in the next section. These functions together
with the scaling functions for the magnetic equation of
state h = t@ s H±(m/rf) and isothermal susceptibility
\t = t ~ 1 x±{ m l T ^) UM are summarized in table ITU

C. Isothermal magnetocaloric effect, A > 3

Before we proceed with the discussion of the peculiar-
ities of the entropic equation of state and of the ther-
modynamic functions following from it, let us introduce
an additional observable - the isothermal magnetocaloric
coefficient. It serves as a direct measure of the heat re-
leased by the system due to the magnetocaloric effect

3 < A < 5

A > 5

f±

±x 2 /2 + x A_1 /4

±x 2 /2 + x 4 /4

H±

^-x x ~ 2 ± X

4

x 3 ± X

\=

1

1

(A-l)(A-2)i A - 3 /4± 1

3x' A ± 1

S

-x 2 /2

-x 2 /2

c±

x 2

x 2

(\-l){\-2)x*- 3 /i± 1

3x' J ± 1

M±

X

X

(\-\)(\-2)x x -3 /4± 1

3x' 2 ± 1

A+/A-

r+/r-

A-3

2

R\

1

1

Rc
Ra

A-5

-J_[^!_l (A-2)(A-3)
A-2 lA-lJ

1/3

TABLE II: Scaling functions and amplitude ratios near the
critical point in scale- free networks. The scaling variable is
x — m/r' 3 . The ratio T + /r~ is taken from Ref. Il3l | and
scaling functions f±, H±, and \± follow from Ref. [lj|.

upon an isothermal increase of the magnetic field and is
defined as (see e.g. [2o| )

M T = —T

dT

(40)

In contradistinction to the heat capacity, which often
does not diverge or is a weakly divergent quantity for
many 3d systems, the magnetocaloric coefficient is fre-
quently strongly divergent at second-order phase transi-
tions [22],[26| and therefore it is instructive to analyze how
this behavior is modified by a scale-free network. Using
Maxwell relations, Mt can be obtained both from the
magnetic or from the entropic equations of state, Eqs.
(jT7j) , (fl~8|) . Therefore an equivalent representation to the
one given in f|40[) representation reads:

M T = -T

,dS
~dH

(41)

Analogous to the first equation in ([7]), we define the di-
mcnsionless isothermal magnetocaloric coefficient as

rriT

Mt
Mb

From the above representation this is

(?s(?ti,t) dm(r,h)

ttit = — (1 ± t)-

dm

01)

(42)

(43)

Recognizing that the last term in (|4"3")) is a dimensionlcss
isothermal susceptibility XT(T,m) and writing it in the
scaling form

XT

m/r

(44)

we arrive at the scaling representation for the dimension-
less isothermal magnetocaloric coefficient rriT

m T = {1±t)t~ IjJ M±{x),

(45)

5

with the scaling function

M±(x) = xx±(x),

(46)

and a scaling relation for the isothermal magnetocaloric
coefficient critical exponent w,

1-/3.

(47)

While the equality (|4T|) is a general one and directly fol-
lows from the scaling form of Eq. (|43| , the relation (|46|)
between functions M±(x) and X±( x ) holds only for sys-
tems where the entropy scaling function has the simple
representation (|27p.

As noticed above, another way to obtain mj is to start
from the magnetic equation of state using the represen-
tation (|4U|) . Then one obtains

TUT = T(l ± T)

dm(r, h)

(48)

Comparing this expression with the formulas obtained
above for the heat capacity (f23|) . (f24)) one arrives at the
relation between the scaling functions M±(x) and C±(x),

C±(x) = xM±(x),
which in particular leads to (c.f. (j46]))
C±(x) = x 2 x±{x).

(49)

(50)

The scaling function Ai±(x) defined above is displayed
for different ranges of the values of A in table |TTJ For the
critical exponents u> we get

A-4
A-3'

3 < A < 5; cj = 1/2, A > 5. (51)

It is easy to find the scaling relation for the critical expo-
nent uj c that governs the field dependence of wit^t = 0, h)

1-/3

LU r =

The values of this exponent read
A-4

(52)

A-2

, 3 < A < 5; cj c = 1/3, A > 5. (53)

Thus while Ch does not diverge (a < 0) for the entire
range 3 < A < 5, my is divergent (u> > 0) over half
that range 4 < A < 5, and is a better locator of the
phase transition there. The above calculated exponents
ui, lu c are displayed together with other exponents in the
comprehensive table U which presents a summary of the
data concerning the temperature and field behavior of
different thermodynamic quantities in the vicinity of the
critical point for different values of A. In the course of the
analysis of different types of critical phenomena in scale-
free networks [![, it has been revealed that the onset of
divergencies of moments of the node-degree distribution

function -P(fc), Eq. ((1) relate to changes in the scaling
scenario of these systems. As one can see from (|5"Tj) , (|53")l
the exponents w and uj c change their sign to become neg-
ative for A < 4: mj- is not divergent at the critical point
any more in the region 3 < A < 4. Therefore, along with
the two marginal values of A = 5 and A = 3 which cor-
respond to the divergencies of (fc 4 ) and (fc 2 ) and define
the 'window' of non-trivial critical behavior on a scale-
free network. The divergence of the third moment of
the node-degree distribution (fc 3 ) leads to a qualitative
change in the critical behavior of the isothermal magne-
tocaloric coefficient.

III. DISCUSSION

As one can see from tablelU the heat capacity exponent
a is negative in the region 3 < A < 5 where a non-trivial
a(X) dependence is observed. This means that the singu-
lar part of the heat capacity Ch vanishes at T c . Taken that
Ch vanishes also at T = and that it is a positive smooth
function of T in between, one concludes that it has a
maximum at some temperature To, where < Tq < T c
for any 3 < A < 5. Therefore the energy fluctuations
are maximal at To (see [l3| for more details). Such be-
havior is a generic feature of systems with a < 0, other
examples include the three dimensional Heisenberg and
planar magnets l27l|. liquid helium [28j |. and disordered
uniaxial magnets [29|. From Eqs. (|3"Tj) and (l3"6l one finds
that c h (T > T c , h = 0) = for any A. This leads to the
following amplitude ratio which holds for all A > 3:

A+/A- = 0.

(54)

Amplitude ratios are known to be universal along with
the scaling functions and critical exponents (see e.g. [§]).
It is appropriate to adduce here how do these ratios
change for systems on scale-free networks. The results
are summarized in the lower part of table [TTJ Besides
than the heat capacity amplitude ratio (|5~4")) , the isother-
mal magnetic-susceptibility amplitude ratio appears to
be A-dependent for 3 < A < 5: r+/T~ = A - 3 [lj].
Using the expressions j[28]). (|3T|) . (j34|) . ((36)) it is straight-
forward to find for the other amplitudes for A > 5

B = D C = 1,

A c = 1/3,

(55)

and for 3 < A < 5,

4 \V(A-3) \-i
B = (- r) , D c = — —

A,

A- 17

1/4 \3/(A-2)

\X-lJ

(56)

A-2'

Now, defining three more amplitude ratios by [1, [H, [l9[

R x = T+D c B 5 -\
Rc = A+T+/B 2 ,
R A = AcD-^+^B- 2 '?,

(57)
(58)
(59)

6

and substituting into these ratios the amplitudes (|55j) and
([55]) . we arrive at their values for the scale-free network,
as listed in table [Til

Let us concentrate now on the scaling functions. As
noted in section UH the entropy scaling function S±(x)
which in the usual Landau theory is given by Eq. (|27j)
keeps its form also in the case of scale-free networks with
3 < A < 5. However, the constant-magnetic-ficld heat
capacity scaling function C± essentially changes in this
region. In Fig. [T]we plot C± as a function of the scaling
variable x = ra/r 13 for different values of A. The most
striking feature in the behavior of the scaling function is
that its asymptotics for large x change for A < 5. Indeed,
for A > 5 the asymptotical value is given by a constant,
C±(x — > oo) = 1/3, whereas in the range 3 < A < 5 the
function behaves as a power law,

C±(x — > oo)

„5-A

(A-l)(A-2)

(60)

In turn, this is reflected in the development of a minimum
in the C_ branch of the function as A decreases (see the
figure) .

a non-zero magnetic field must increase the value of m
one arrives at the following minimal values of the scaling
variable x:

^min
Xmin

1, A > 5,
4 \ V(>-3)

A- 1

(61)

3 < A < 5. (62)

Therefore, the curves for the scaling function C_ in Fig.
[1] originate at the corresponding minimal values of x de-
fined by the relations (JBTJ), ([62]l.

In turn, as explained in section [Hi one may re-express
the scaling function C± in terms of the scaled magnetic
field y using Eq. ([37]) . Corresponding plots for the scal-
ing function C±(y) in this variable are given in Fig. [2]
for different values of A. Again, one observes a change in
the asymptotics of the scaling function: instead of a con-
stant at A > 5, for 3 < A < 5 the asymptotic functional
dependence is given by

C±{y -> oo) =

1

(A-2)V(A-1)

7/A-2

(63)

y

FIG. 1: Heat capacity scaling functions C-(x) (dotted curves,
blue online) and C+(x) (solid curves, red online) as functions
of the scaling variable x = m/r' 3 at A > 5, A = 4.8, and A =
4.5 (lower, middle and upper pairs of curves, respectively).

Another particular feature of the plots of Fig. Q] is
inherent to the presentation of the scaling plots in the
C_-a;-plane and is connected to the presence of a pole
in C±(x) for small x. As one sees immediately from
Eqs. (|3Tj) and (|36|) . this pole occurs at x = l/\/3 and

x = (4/[(A - 1)(A - 2)]) 1/(A ~ 3) for A = 5 and 3 < A < 5,
correspondingly. However, the physical values of the scal-
ing variable x do not cover the region where the pole
occurs. Indeed, from the magnetic equations of state
(|28p. (f34| one may obtain the solutions for the magne-
tization at zero magnetic field m(r,h = 0). Taking that

FIG. 2: Heat capacity scaling functions C-(y) (dotted curves,
blue online) and C+(y) (solid curves, red online) as functions
of the scaling variable y = h/r^S at A > 5, A = 4.8, and A =
4.5 (lower, middle and upper pairs of curves, respectively).

In figures [3] and [4] we give the plots of the scaling func-
tions A4± for the isothermal magnetocaloric coefficient in
the scaling variables x = m/r 13 and y = h/r^ s , respec-
tively. As discussed at the end of the previous section,
tut changes its behavior at A = 4. This feature is also
reflected in the behavior of the scaling functions: their
asymptotics change at A = 4. Indeed, for A > 5 the func-
tion decays as J\A±(x — > oo) ~ l/3x whereas from the

7

FIG. 3: Scaling functions fo the isothermal magnetocaloric
coefficient. M-(x) (dotted curves, blue online) and M+(x)
(solid curves, red online) as functions of the scaling variable
x — m/r 13 at A > 5, A = 4, and A = 3.8 (lower, middle
and upper pairs of curves, respectively). The decay which is
observed for A > 4 switches to power law growth for A > 4.
The functions approach the constant value M± — 2/3 for
A = 4.

asymptotic behavior in the region 3 < A < 5,

M ^^^= (X-1) { X-2) X ^ (64)

one concludes, that for A < 4 the power law decay
switches to a power law growth, while M±(x — >• oo) =
const for the marginal value A = 4. The corresponding
asymptotics in the variables y is of the form

-M±(y^oo) = — L, A>5, (65)

M ±to^°°) = (^((A^T))"^'
3 < A < 5.

IV. CONCLUSIONS

Usually the universality of critical phenomena is at-
tributed to the presence of only a few relevant global pa-
rameters. Different systems that share the same values
of these global variables manifest the same criticality. A
classical example is given by the famous 3d Ising model
universality class that is inherent to the critical behav-
ior of such differing systems as uniaxial magnets, simple

FIG. 4: Isothermal magnetocaloric coefficient scaling func-
tions M-(y) (dotted curves, blue online) and M+(y) (solid
curves, red online) as functions of the scaling variable y =
m/r 136 at A > 5, A = 4, and A = 3.8 (lower, middle and upper
pairs of curves, respectively).

fluids or binary alloys. The critical behavior in all these
systems is quantitatively described by the same values
of the critical exponents, amplitude ratios, and by the
same form of the scaling functions. As this paper demon-
strates, in particular the usual 'Euclidean space' un-
derstanding of universality of critical phenomena breaks
down if the critical behavior occurs on a scale-free net-
work. The presence of high-degree vertices (hubs) may
lead to substantial changes in ordering processes. The
parameter that controls the 'importance' of the hubs is
the node-degree distribution exponent A, Eq. ((T]), and it
is this parameter that plays the role of a global variable
as far as the critical behavior is considered.

In particular, the universal quantities that govern crit-
icality become A-dependcnt for small enough A and in
this way the network structure is felt. However, the pres-
ence of magnetization is necessary to 'feel' the network
structure. To give an example, the structure matters for
T < T c at any h and for T > T c for h ^ (c.f. that
amplitude T_ is A-dcpendcnt, whereas T + is not). An-
other interesting observation is that the fluctuation in
network structure only enters via the magnetization and
since the entropy S is measured at constant magnetiza-
tion, it is given by a usual Landau-like mean field value
for any A > 3. This makes a difference between the global
parameters A and dimension d regarding its influence of
the fluctuations on calculating the singular behaviour of
the physical quantities: no renormalization-group calcu-
lation is necessary.

8

In this paper we completed the quantitative descrip-
tion of critical behavior in scale-free networks by calcu-
lating the entropic equation of state, the resulting scaling
functions as well as the universal amplitude ratios. The
corresponding formulas, together with other data for the
critical exponents and amplitudes are summarized in Ta-
bles U and [TTJ They constitute a comprehensive list of ob-
scrvables that describe the scaling and characterize the
criticality in scale-free networks.

The starting point for our study was the asymptotic
form of the free energy in the vicinity of a critical point in
a scale- free network, Eqs. (O, @. The validity of this ex-
pression has been proven at different levels of rigor using
microscopic approaches based on the recursion relations
(lo| the replica method [|| or phenomenological Landau
approaches [ll|, as well as mean field theory [l2[ . It is

instructive to note here that, because the networks under
discussion have a local tree-like structure, the mean-field
approximation is asymptotically exact. One of the con-
sequences of this fact is that the values of the exponents
do not change if an Q(m)-symmetrical order parameter
is considered (see e.g. |L3[), as it is usual in the Landau
theory.

Acknowledgement

This work was supported by the Austrian Fonds
zur Forderung der wissenschaftlichen Forschung under
Project No. P19583-N20 and by the ARF Scheme of
Coventry University.

[1] S. N. Dorogovtsev and A. V. Goltsev, Rev. Mod. Phys.
80, 1275 (2008).

[2] S. Galam, Physica A 274, 132 (1999); Int. J. Mod. Phys.
C 19, 409 (2008); K. Sznajd-Weron and J. Sznajd, ibid.
11, 1157 (2000); K. Sznajd-Weron, Acta Phys. Pol. B
36, 2537 (2005); D. Stauffer and S. Solomon, Eur. Phys.
J. B 5 7, 473 (2007); K. Kulakowski and M. Nawojczyk,
e-print larXiv:0805.3886l

[3] B. Tadic, K. Malarz, and K. Kulakowski, Phys. Rev. Lett.
94, 137204 (2005).

[4] See, e.g., H. E. Stanley, Introduction to Phase Transi-
tions and Critical Phenomena (Oxford University Press,
Oxford, 1971); C. Domb, The Critical Point (Taylor &
Francis, London, 1996).

[5] If the Helmholtz potential F(r, m)is a generalized homo-
geneous function, one can show [f| that the other ther-
modynamic potentials are also generalized homogeneous
functions.

[6] A. Hankey, H. E. Stanley, Phys. Rev. B 6, 3515 (1972).
[7] C. Domb and D. L. Hunter, Proc. Phys. Soc. 86, 1147
(1965); R. B. Griffiths, Phys. Rev. 158, 176 (1967);

A. Z. Patashinskii and V. L. Pokrovskii, Zh. Eksp. Teor.
Fiz. 50, 439 (1966) [Sov. Phys. JETP 23, 292 (1966)];

B. Widom, J. Chem. Phys. 43, 3898 (1965).

[8] See e.g. V. Privman, P. C. Hohenberg, and A. Aharony,

in: Phase Transitions and Critical Phenomena (Edited

by C. Domb, J. L. Lebowitz), 14, 1-134 (1991).
[9] M. Leone, A. Vazquez, A. Vespignani, and R. Zecchina,

Eur. Phys. J. B 28, 191 (2002).
[10] S. N. Dorogovtsev, A. V. Goltsev, and J. F. F. Mendes,

Phys. Rev. E 66, 016104 (2002).
[11] A. V. Goltsev, S. N. Dorogovtsev, and J. F. F. Mendes,

Phys. Rev. E 67, 026123 (2003).
[12] F. Igloi and L. Turban, Phys. Rev. E 66, 036140 (2002).
[13] V. Palchykov, C. von Ferber, R. Folk, and Yu. Holovatch,

Phys. Rev. E 80, 011108 (2009).
[14] V. Palchykov, C. von Ferber, R. Folk, Yu. Holovatch, and

R. Kenna. Phys. Rev. E 82, 011145 (2010).
[15] D.J. Amit. Field Theory, the Renormalization Group,

and Critical Phenomena (World Scientific, Singapore,

1989).

[16] R. Krasnow and H.E. Stanley. Phys. Rev. B 8, 332

(1973) .

[17] D.S. Simons and M.B. Salamon. Phys. Rev. B 10, 4680

(1974) .

[18] M. Barmatz, P.C. Hohenberg, and A. Kornblit. Phys.

Rev. B 12, 1947 (1975).
[19] A. Aharony and P.C. Hohenberg. Phys. Rev. B 13, 3081

(1976).

[20] M.B. Salamon, S.E. Inderhees, J.P. Rice, B.G. Pazol,
D.M. Ginsberg, and N. Goldenfeld. Phys. Rev. B 38, 885
(1988).

[21] B. Stroka, J. Woznitza, E. Scheer, H. v. Lohneysen, W.

Park, and K. Fischer. Z. Phys. B 89, 39 (1992).
[22] T. Plackowski and D. Kaczorowski. Phys. Rev. B 72,

224407 (2005).

[23] V. Franco, A. Conde, J.W. Romero-Enrique, Y.I.
Spichkin, and V.I. Zverev, A.M. Tishin. J. Appl. Phys.
106, 103911 (2009).

[24] F. J. Wegner, in Phase Transitions and Critical Phe-
nomena, edited by C. Domb and M. S. Green (Aca-
demic Press, London, 1976), Vol. VI, p. 8; R. Kenna,
D. A. Johnston, and W. Janke, Phys. Rev. Lett. 96,
115701 (2006); Phys. Rev. Lett. 97, 155702 (2006), Phys.
Rev. Lett. 97, 169901(E) (2006).

[25] Entropies at constant magnetic field and at constant
magnetization coincide.

[26] V.K. Pecharsky, K.A. Gschneidner Jr. J. Mag. Mag. Mat.
200, 44 (1999); T. Plackowski, A. Junod, F. Bouquet, I.
Sheikin, Y. Wang, A. Jezowski, and K. Mattenberger.
Phys. Rev. B 67, 184406 (2003); T. Plackowski, D. Kac-
zorowski, and Z. Bukowski. Phys. Rev. B 72, 184418
(2005).

[27] D. L. Connelly, J. S. Loomis, and D. E. Mapother. Phys.

Rev. B 3, 924 (1971); F. L. Lederman, M. B. Salamon,

and L. W. Shacklette. Phys. Rev. B 9, 2981 (1974).
[28] J. A. Lipa, D. R. Swanson, T. C. P. Chui, and U. E.

Israelsson. Phys. Rev. Lett. 76, 944 (1996).
[29] R. Folk, Yu. Holovatch, and T. Yavors'kii. Uspiekhi

Fizichieskikh Nauk 173, 175 (2003) [Physics-Uspiekhi,

46, 169 (2003)].